Future climate projections for South Aegean and Salzburg region based on statistical and dynamical downscaling - similarities and discrepancies

Abstract As climate change is advancing at a quickened pace, the need for a better understanding of local climate change impacts is increasing, especially since the new generation of general circulation models (CMIP6) displays a (statistically nonsignificant) higher global sensitivity than previous climate projections. Within the presented study, two CMIP6 general circulation models are statistically and dynamically downscaled for the Salzburg and South Aegean region to investigate local future climate states under two different shared-socioeconomic pathways until 2100. Comparisons of both methods have been performed previously, however, within this study the innovation lies in the detailed investigation of two regions representing different climatic conditions, geography, and the availability of observational data. Furthermore, the dynamical downscaling is performed down to convection permitting scale, therefore providing information on a 4 km 2 grid and enabling the representation of local features, as is the case when training a statistical model with observations. The chosen set-up enables us to ask the following questions (i) how do the two methodologies compare for the past period (1981–2010) against high-resolution reanalysis data, (ii) does their skill depend on the region investigated (South Aegean/Salzburg) and on the spatial resolution of the dynamical downscaled data, (iii) how do future conditions coincide/differ between the applied methods and regions?. The results display that, if observation data is available, temperature is better simulated for the past period using statistical downscaling, while precipitation patterns are more realistically met by the dynamical models. Concerning future climate change, the impact is more severe according to the dynamical models in the South Aegean region, while for Salzburg the statistical models display the stronger change, indicating the importance and uncertainty of the chosen downscaling method on localized results.